Catalog, Lifestyle and Close-Up: Three Backgrounds for One Product

Three backgrounds, one product: the setting changes, the item does not. Its material, color and proportions stay fixed from image to image. A text prompt alone is fine for exploring an idea. For a shop listing, start from your own product reference and confirm that every output shows the item as it really is.

Model and price

This works with GPT Image 2.5, currently $0.05 per image. Confirm the cost of your settings on the pricing page or in the calculator.

• No subscription, and the balance has no expiry date.
• Failed generations cost nothing.
• Successful images are billed whether you use them or not.

Write the fixed part of the brief first

Before any background, list what must never change:

• Color and finish
• Parts (handles, lids, straps, closures)
• Texture and material
• Logo placement
• Shape and proportions

Paste that list into every prompt in the series.

The three variants

1. Catalog: plain background, soft shadow, crisp edges.
2. Lifestyle: a real setting with one object that shows scale. Do not add props that a buyer could take as part of the package.
3. Detail: close view of material or a feature. Reject any made-up quality badges or labels the model invents.

A starting request

The body for POST /v1/generate. When you use a reference, give a URL the API can open.

{
  "model": "gpt-image-2.5",
  "input": {
    "prompt": "A matte enamel water bottle on a smooth warm gray backdrop, gentle shadow, catalog product shot, no branding, no text"
  }
}

Check the set side by side

Put all three images next to each other and look for:

• Color drift compared with the real product.
• Geometry changes in handles, caps and closures.
• Misleading extras that suggest accessories are included.
• A clear job for each image: listing thumbnail, mood shot, detail.

Then export everything at the same aspect ratio and write alt text that names the product and the view.

Questions

Can I list a product using only text-generated images?

Not without checking them against the real item. A prompt alone does not describe what you are selling, and the model can add details that do not exist.

How do I know the cost before generating?

Studio shows the price before you send, and the API response carries it in estimated_cost. Look at the model schema too.

Which inputs does GPT Image 2.5 accept?

See the GPT Image 2.5 page and the generate docs.